Considerable research demonstrates a strong tendency to use facial appearance when forming first impressions. Moreover, these impressions show remarkable consensus, yielding significant social consequences. The long-range objective of the proposed research is to explain consensual first impressions of faces and to develop methods for ameliorating their negative social consequences. The working hypothesis is that the psychological qualities that are accurately revealed by the functionally significant facial qualities that mark babies, unfitness, emotion, or identity are overgeneralized to people whose facial structure resembles that of babies, a particular level of fitness, a particular emotion, or a particular identity. The research has three specific aims. One is to use connectionist modeling to test the facial identity overgeneralization hypothesis that the tendency for responses to strangers to vary with their facial resemblance to known individuals contributes to racial prejudice and stereotyping. The connectionist modeling experiments seek to demonstrate that the physical similarity between two faces can in and of itself account for similar impressions of them quite apart from similarities in the social categories of the faces. The second aim is to test whether generalized mere exposure effects can be used to reduce race and age prejudice and stereotyping, as predicted by the facial identity overgeneralization hypothesis. The mere exposure experiments seek to demonstrate that increasing the familiarity of an out-group facial prototype will decrease negative reactions to out-group members. The third aim is to use functional Magnetic Resonance Imaging (fMRI) to investigate neural activation patterns in response to faces that are predicted from each of the three overgeneralization hypotheses. The fMRI experiments seek to determine whether categories of faces that are differentiated by human judges' ratings and by the activation they elicit in connectionist modeling experiments also elicit distinct patterns of neural activation, thereby demonstrating a neural substrate for the overgeneralization effects. By focusing on the structured facial information that influences prejudice and stereotypes, the proposed research brings a novel theoretical perspective to the field of social cognition, demonstrating that the intrinsic properties of faces make a significant contribution to social biases that have been largely viewed as social constructions. It also suggests novel interventions for reducing prejudice.